基于时频卷积神经网络的供水管道漏损识别

赖凌轩,柳景青,周一粟,李秀娟

Identification of leakage in water supply pipelines based on time-frequency convolutional neural network

Lingxuan LAI,Jingqing LIU,Yisu ZHOU,Xiujuan LI

表 6 不同分类模型的性能比较

Tab.6 Performance comparison of different classification models

分类模型 输入特征 Acc F1 无漏损 漏损 噪声 低压 中压 高压 STFT-CNN 时频谱图 0.952 0.972 0.988 0.987 0.924 0.929 0.935 MFCC-CNN MFCC 0.859 0.960 0.975 0.957 0.759 0.791 0.785 DT STD、RMS、ZCR、PSD 0.684 0.868 0.920 0.841 0.458 0.571 0.594 SVM ApEn, MFCC, IMF 0.844 0.944 0.968 0.967 0.720 0.764 0.775 KNN 一维时序信号 0.831 0.732 0.884 0.854 0.873 0.875 0.901 XGBoost 一维时序信号 0.763 0.900 0.948 0.918 0.577 0.661 0.680